Can Species Distributions Models Help to Design Conservation Strategies for Narrow-Ranged Species under Climate Change? A Case Study from Santolina Genus †

Climate change is dramatically threatening biodiversity. Narrowly distributed species are especially exposed to extinction risk due to their narrow ecological niche. We used Species Distribution Models at fine spatial resolution (50 m) to investigate changes in the distribution of three range-restricted species of Santolina (Asteraceae) endemic to the Mediterranean Basin (S. decumbens, S. ligustica, S. pinnata). We assessed the future potential range under an optimistic and a pessimistic scenario, and analysed distribution change taking into account three different areas: the distributional range (calculated as convex hull), an area 5 km larger than the distributional range, and a buffer (1 km) around occurrences. Santolina ligustica is expected to dramatically reduce its range under both scenarios, S. decumbens is expected to increase its range under both scenarios and S. pinnata is expected to dramatically reduce its range under pessimistic scenario and to increase it under optimistic one. Moreover, under the optimistic scenario, S. ligustica and S. pinnata show a very high range loss in all areas but the range gain is major in the largest area than in the other two areas. This result suggests that, in the future, suitable areas will occur mainly outside of the current distributional range and that assisted colonization might be necessary to assure species survival. Differently the third species has a lower range loss and higher range gain within the distributional range and in the buffer around occurrences, suggesting the possibility of survival in microrefugia within its distributional range despite a wide reduction in suitable habitat. These results might help to design strategies for species conservation in face of future climate change.


Introduction
Climate change is increasingly affecting the distribution of species, particularly those who have a narrow distributional range [1][2][3][4]. These small-ranging species are in double jeopardy because they have a great intrinsic vulnerability to stochastic disturbances due to their low population size and a limited ability to escape warming because of poor dispersal capability [5]. Consequently, in situ protection of natural habitats may be inadequate to reduce the loss of populations in narrow range species because of the velocity of human-induced climate change [6]. Therefore, it may be necessary to move individuals into new suitable areas, likely not reachable by endemic taxa due to their dispersal limitations. However, the survival rate of translocated plants is generally low mainly because of the release in unsuitable habitat [7,8]. Thus, to conserve small-ranging species it is necessary not only to assess their vulnerability to climate warming, but also to understand where suitable areas will occur in the future in order to detect optimal locations for reintroductions and to maximize the success rate.
Nowadays species distribution models (SDMs) are a broadly used statistical tool to forecast the impact of climate change on biodiversity [9]. Nevertheless, until now their applications to translocation planning have been scarce.
The Mediterranean Basin is one of the main world biodiversity hotspots and it is particularly rich in endemic species [10,11]. Santolina genus (Asteraceae) is widespread throughout the Mediterranean Sea. It includes 24 species, several of which are endemic. In this study we selected S. decumbens Mill., S. ligustica Arrigoni and S. pinnata Viv., three species with a narrow distribution range, endemic to the North Tyrrhenian area from Provence to Tuscany.
We used SDMs to analyze the potential effects of climate change on these three narrow range species of Santolina. To assess the possibility of assisted translocation, we evaluated the occurrence of suitable areas at different distance from the currently known distribution of the species.

Materials and Methods
The three selected Santolina species are characterized by a relatively high number of occurrences in a restricted area. Santolina decumbens is endemic to Southern France, located in the South East of France (Provence), S. ligustica and S. pinnata are Italian endemisms, located in the North West of Italy (Liguria and Tuscany). Santolina decumbens grows in dry fields and garrigue on basic rocky soils, alluvial deposits, marls and silty soil between 10 and 1025 m a.s.l. [12]. Santolina ligustica grows in garrigue, slopes and rocky fields on ophiolitic substrates between the sea and 650 m a.s.l. [13]. Santolina pinnata grows on sunny cliffs, rocks and limestone soils between 500 and 1500 m a.s.l. [14][15][16].
Occurrence data were obtained from local database [17] and from the project "PLAN.T.S. 2.0-towards a renaissance of PLANt Taxonomy and Systematics". We took into account only occurrences detected using a GPS tracker.
We downloaded 19 bioclimatic variables both for the present and the future (2080) from CHELSA climate database website [18] at of 30-s (~1 km) spatial resolution. We selected, among those available, the most optimistic scenario-named rcp26-and the most pessimistic one-named rcp85. To take into account microclimatic conditions [19], we statistically downscaled bioclimatic variables at 50 m spatial resolution using a physiographically informed model fitted with a geographically weighted regression (GWR). We used the following physiographic variables derived from a digital elevation model downloaded from the CGIAR-CSI GeoPortal website [20]: distance of the sea, slope, north exposition and south exposition. We then used the first two axes of the Principal Component Analysis (PCA) as environmental variables for species distribution modelling in order reduce the transferability issue [21]. We calculated the PCA using "ade4" package implemented in R [22].
We converted the continuous probability values into binary presence-absence form. To reduce the possibility of prediction bias linked by the choice of the threshold, we used three different thresholds implemented in the PresenceAbsence package [34], performing equally or better than others [35]. Following the majority consensus rule, we considered the species as occurring in a cell only if the model predicts at least 50% of its presence there.
To assess the position of possible area for translocation in relationship to the species distribution, we considered three different ranges: the distributional range (calculated as convex hull), an area 5 km larger than the distributional areas, and a buffer (1 km) around occurrences. For all these three cases, we calculated the number of currently suitable cells. Moreover, we calculated: the number of cells that will remain suitable (range stable); the number of cells that will no longer be suitable (range loss); and the number of cells that are not currently suitable but that will be suitable under future climate (range gain). Lastly, we calculated range change, which is based on the number of potential cells gained or lost [36].

Results and discussion
According to the thresholds established from Araujo et al. [37], under current climate scenario, all modelling algorithms had a good model performance ranging from 0.6 to 0.95.

Future Impacts of Climate Change on the Distribution of Santolina Species
The three species are projected to have a range loss higher in pessimistic than in optimistic scenario, even if they will be probably affected differently by climate change (Tables 1 and 2). In fact, S. decumbens will probably increase its distributional range under the optimistic scenario and will moderately contract it under pessimistic one. Santolina pinnata will expand its distributional range under the optimistic scenario and will lose all the climatically suitable areas under the pessimistic. Lastly, S. ligustica will lose a high percentage of its current potential range under both scenarios (Tables 1 and 2). Table 1. Percentage of range loss (RL), range gain (RG) and range change (RC) under future optimistic scenario for the three studied taxa. For each species, range dynamic was calculated in three different areas: the distributional range (DR), an area 5 km larger than the distributional range (wDR), and a 1 km buffer around occurrences (Bf). The difference between the two scenarios is in line with the observation that climate projected under optimistic scenario fall within the climatic variability that species already experienced during the Holocene [38,39]. Despite range contraction seems to be the main trend under future climate [40], potential range expansion under a moderate global warming, as observed in S. decumbens, was previously observed in other Mediterranean thermophilus species [41,42]. The different trends detected in the three species is in line with the idea that species sensitivity to climate change is affected by niche properties and the difference between current and future climate in which species grow [43]. In fact, the two species with the narrow altitudinal range (i.e., S. ligustica and S. pinnata) seem to be more threatened by climate change than S. decumbens which has a wider altitudinal range. This is likely due to the temperature increase which will remain within the climatic conditions already experienced by populations at lower altitudes. Moreover, under the pessimistic scenario we projected a lower extinction risk in S. decumbens than detected by Dagnino et al. [43], using in our study a finer resolution. This result supports the idea that local microrefugia may play a role in holdout to species extinction under future climate change [44]. Moreover, as already suggested by previous studies [44,45], S. ligustica may be also threatened by urbanization besides climate change, further increasing its risk of extinction.

Differences in Range Change among Areas
Under the optimistic scenario, in S. ligustica and S. pinnata the range gain is higher in the largest areas (i.e., 145% and 128% vs. 16% and 50%) differently to S. decumbens, which will gain more suitable cells in the two narrow areas (i.e., 61% and 65% vs. 46%; Table 1). Similarly, under pessimistic scenario the range loss in S. ligustica and S. pinnata is higher in all the three areas, while S. decumbens will loss part of suitable range considering the wDR area but will gain suitable range considering the Bf area (Table 2). This result suggests that future suitable areas for S. ligustica and S. pinnata would not be close to the current occurrences, differently to S. decumbens. The ability of species to cope with the climate change is affected by the possibility to reach new suitable areas that is, in turn, related to the location of suitable areas and dispersal capability [40]. Santolina has a poor dispersal capability having seeds with a low efficient plume [46]. Consequently, S. ligustica and S. pinnata, the two species for which future suitable areas would not be close to the current occurrences, might not be able to migrate fast enough to counteract climate change. For these species, the assisted colonization should be considered as a proactive conservation activity [47]. Differently, in S. decumbens, which is projected to have suitable cells near the occurrences, conservation strategy may be built around microrefugia (both ephemeral and stable) and stepping-stones [48,49].

Conclusions
The reduction forecasted for the three species underlines the serious potential impact of climate change in the north Mediterranean area. Taken together, our results suggest that some Mediterranean species may generally be favored by future climate change. This is the case of S. decumbens, the species with the widest altitudinal range, which would shift its distributional range. The other two species (S. ligustica and S. pinnata), with a narrow altitudinal range, would be not able to adjust their geographic distribution appropriately, and, in turn, would be more affected by the climate change. In S. pinnata the persistence of suitable area close to the current distribution would allow in situ conservation and translocation. Differently in S. ligustica, for which future suitable areas would be outside of the current distributional range, it would be necessary to carry out assisted colonization. Our study underlines the importance of SDMs as tool to design strategies for species conservation in face of future climate change. SDMs may provide a cost-benefit tool for planning assisted colonization and conservation translocation.